RSSL (version 0.9.7)

SelfLearning: Self-Learning approach to Semi-supervised Learning

Description

Use self-learning (also known as Yarowsky's algorithm or pseudo-labeling) to turn any supervised classifier into a semi-supervised method by iteratively labeling the unlabeled objects and adding these predictions to the set of labeled objects until the classifier converges.

Usage

SelfLearning(X, y, X_u = NULL, method, prob = FALSE, cautious = FALSE,
  max_iter = 100, ...)

Arguments

X

matrix; Design matrix for labeled data

y

factor or integer vector; Label vector

X_u

matrix; Design matrix for unlabeled data

method

Supervised classifier to use. Any function that accepts as its first argument a design matrix X and as its second argument a vector of labels y and that has a predict method.

prob

Not used

cautious

Not used

max_iter

integer; Maximum number of iterations

...

additional arguments to be passed to method

References

McLachlan, G.J., 1975. Iterative Reclassification Procedure for Constructing an Asymptotically Optimal Rule of Allocation in Discriminant Analysis. Journal of the American Statistical Association, 70(350), pp.365-369.

Yarowsky, D., 1995. Unsupervised word sense disambiguation rivaling supervised methods. Proceedings of the 33rd annual meeting on Association for Computational Linguistics, pp.189-196.

See Also

Other RSSL classifiers: EMLeastSquaresClassifier, EMLinearDiscriminantClassifier, GRFClassifier, ICLeastSquaresClassifier, ICLinearDiscriminantClassifier, KernelLeastSquaresClassifier, LaplacianKernelLeastSquaresClassifier(), LaplacianSVM, LeastSquaresClassifier, LinearDiscriminantClassifier, LinearSVM, LinearTSVM(), LogisticLossClassifier, LogisticRegression, MCLinearDiscriminantClassifier, MCNearestMeanClassifier, MCPLDA, MajorityClassClassifier, NearestMeanClassifier, QuadraticDiscriminantClassifier, S4VM, SVM, TSVM, USMLeastSquaresClassifier, WellSVM, svmlin()

Examples

data(testdata)
t_self <- SelfLearning(testdata$X,testdata$y,testdata$X_u,method=NearestMeanClassifier)
t_sup <- NearestMeanClassifier(testdata$X,testdata$y)
# Classification Error
1-mean(predict(t_self, testdata$X_test)==testdata$y_test) 
1-mean(predict(t_sup, testdata$X_test)==testdata$y_test) 
loss(t_self, testdata$X_test, testdata$y_test)